Add utility to convert inhouse 10x data to adata format

[sgratiy]

In house data uses a form of 10x-Genomics-formatted hdf5 while we only support AnnData.

The basic code snippet doing the conversion is residing here:
http://confluence.corp.alleninstitute.org/display/IT/Transcriptomic+Clustering+Pipelines+with+Test+Datasets

We need to have a robust and easily available utility for converting 10X to AnnData in CSR format

• Add conversion script utility
• Handle exception for the unexpected/missing dataset names.
• Handle possible MemoryError that could result from loading large dataset.
• Check for max value of expression, if possible use uint16 to store counts

Validation:

• Conversion module in utils.py
• An executable running from the command line with progress report/expected conversion time/actual time and memory usage.
• show that both files include the same dataset sizes using h5ls or h5dump